Manual Therapy 18 (2013) 124e129

Contents lists available at SciVerse ScienceDirect

Manual Therapy journal homepage: www.elsevier.com/math

Original article

The predictive validity of the Örebro Musculoskeletal Pain Questionnaire and the clinicians’ prognostic assessment following manual therapy treatment of patients with LBP and neck painq H. Dagfinrud a, b, *, K. Storheim c, b, g, L.H. Magnussen d, e, T. Ødegaard e, I. Hoftaniska e, L.G. Larsen e, P.O. Ringstad e, F. Hatlebrekke f, M. Grotle a, c a

Dept of Rheumatology, Diakonhjemmet Hospital, Diakonveien 12, 0370 Oslo, Norway University of Oslo, Norway FORMI, Oslo University Hospital, Norway d Dept of Health and Social Sciences, Bergen University College, Norway e Dept of Public Health and Primary Health Care, University of Bergen, Norway f Hans & Olaf Fysioterapi, Oslo, Norway g Orthopaedic Department, Oslo University Hospital, Norway b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 28 December 2011 Received in revised form 22 August 2012 Accepted 30 August 2012

The purpose of this study was to compare the predictive ability of the standardised screening tool Örebro Musculoskeletal Pain Questionnaire (ÖMPQ) and the clinicians’ prognostic assessment in identifying patients with low back pain (LBP) and neck pain at risk for persistent pain and disability at eight weeks follow-up. Patients seeking care for LBP or neck pain were recruited by 19 manual therapists in Norway. Patients completed the ÖMPQ and the low back- or neck specific Oswestry Disability Index/Neck Disability Index at baseline and 8 weeks after first consultation. The manual therapists filled in their assessment of patient’s prognosis immediately after the first consultation, blinded for patient’s answers to the questionnaire. A total of 157 patients (81with neck pain and 76 with LBP) were included. The best odds for predicting the outcome for LBP patients was found for the clinicians’ assessment of prognosis (LRþ ¼ 2.1 and LR  ¼ 0.55), whereas the likelihood ratios were similar for the two tools in the neck group. For LBP patients, both the clinicians’ assessment and the ÖMPQ contributed significantly in the separate regression models (p ¼ 0.02 and p ¼ 0.002, resp), whereas none of the tools where significant contributors for neck patients (p ¼ 0.67 and 0.07). Neither of the two methods showed high precision in their predictions of follow-up at eight weeks. However, for LBP patients, the ÖMPQ and the clinicians’ prognostic assessment contributed significantly in the prediction of functional outcome 8 weeks after the initial assessment of manual therapist, whereas the prediction for neck patients was unsure. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Manual therapy Prediction Prognostic assessment Yellow flags

1. Introduction Manual therapy, in terms of mobilisation techniques and spinal manipulation, is a frequently used treatment for patients with low back pain (LBP) and/or neck pain seeking help in the primary health q The work should be attributed to: Diakonhjemmet Hospital, Diakonveien 12, 0370 Oslo, Norway. * Corresponding author. Dept of Rheumatology, Diakonhjemmet Hospital, Diakonveien 12, 0370 Oslo, Norway. Tel.: þ47 41906633. E-mail address: h.s.dagfi[email protected] (H. Dagfinrud). 1356-689X/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.math.2012.08.002

care sector. A minor proportion of approximately 5e10% is at risk for developing persistent problems, but enormous costs are generated from this small proportion (Nachemson and Jonsson, 2000; Sattelmayer et al., 2011). Early detection of patients at risk for developing persistent problems could optimise treatment decisions and reduce burden of disease for the individual patient and for society (Sattelmayer et al., 2011). Psychosocial factors, collectively termed “yellow flags”, are important risk factors for long-term disability and guidelines for the management of LBP emphasise the importance of screening for psychosocial factors (van Tulder et al., 2006; Airaksinen et al., 2006).

H. Dagfinrud et al. / Manual Therapy 18 (2013) 124e129

A prognostic assessment of the disease course is part of a clinician’s clinical reasoning, although it is not explicitly formulated. A standardised assessment can be helpful in making clinical decisions regarding prognosis and treatment allocation. Such assessment could be done by simply reporting the clinicians’ global impression of whether the patient is likely to improve or not, or by using a patient-reported standardised screening tool, like for example the Örebro Musculoskeletal Pain Questionnaire (ÖMPQ). The ÖMPQ is a thoroughly tested self-administered screening tool, developed to identify patients with acute or subacute musculoskeletal pain who are at risk of persistent pain and disability (Linton and Boersma, 2003; Hockings et al., 2008). Comprehensive questionnaires may be time-consuming to administer and interpret in a busy clinical practice. As far as we know, there has been no attempt to compare the use of the ÖMPQ with the clinicians’ assessment of prognosis. Thus, the purpose of this study was to compare the predictive ability of the ÖMPQ and the clinicians’ prognostic assessment in identifying patients with LBP and neck pain at risk for persistent pain and disability at eight weeks follow-up. 2. Methods and materials 2.1. Subjects and design The study is a prospective cohort study with follow-up 8 weeks after the initial consultation with a manual therapist. A convenient sample of patients was recruited over a period of six months by 19 manual therapists from three different counties in Norway. Only patients with low back- or neck pain seeking the manual therapist directly, without referral from a general practitioner, were asked to participate in this study. Further, patients had to be aged 18 years or more, and not been treated for low back- or neck pain during a period of 4 weeks before enrolment. Exclusion criteria were pregnancy, not understanding Norwegian language and abuse of drugs or alcohol. The study was carried out according to the Helsinki Declaration and was approved by the regional medical research ethics committee. Written informed consents were obtained from all patients. 2.2. Treatment The patients in this study received manual therapy, which is a concept including hands-on joint and muscle mobilisation techniques, information and exercise modalities. The doses of the different elements and the frequency of the delivery were individually adapted, in line with current practice. 2.3. Measurement procedure

125

collected at baseline only. Back- or neck specific daily functioning (Oswestry Disability Index (ODI)/Neck Disability Index (NDI)) was collected at baseline and re-administered at follow-up 8 weeks later (dependent variable). The ÖMPQ contains 25 items in which 21 (items 5e25) are included in a sum score. The scored-items assess pain, previous sick leave, anxiety and depression, activity limitations, coping, work characteristics/-satisfaction, fear-avoidance beliefs, and patient’s expectations to improve. The scored-items are summed to provide a total score, range 0e210, with higher scores indicating a higher risk of a poor outcome. Different cut-off values have been suggested in various studies (Sattelmayer et al., 2011). In this study, the ÖMPQ was categorised into three groups: patients with low risk for prolonged disability ( 105 (Linton and Hallden, 1998). The reliability of the Norwegian and Swedish version of the ÖMPQ has been reported to be good (Linton and Hallden, 1998; Grotle et al., 2006). Functional limitations due to back- or neck pain were assessed by the ODI or the NDI, respectively. The Oswestry Disability Index, version 2.0 (ODI 2.0) is a frequently used and recommended outcome measure within back research (Baker et al., 1990; Deyo et al., 1998; Grotle et al., 2003). Version 2.0, which is recommended by the original authors, is a modification of the original ODI (Baker et al., 1990). The patients rate their perceived disability regarding 10 different items (pain intensity, personal hygiene, lifting, walking, sitting, standing, sleeping, sexual activity, social activity, and travelling) on a 6-point scale, with 0 representing no limitation and 5 representing maximal limitation. A percentage score from 0 to 100 is calculated (100 representing maximal limitation). Version 2.0 used in this study has been translated and cross-culturally adapted for Norwegian patients (Grotle et al., 2010). A minimal clinically important change (MCIC) for the ODI is reported to be 10 points (Ostelo et al., 2008). The NDI is a modification of the ODI version 1.0 (Fairbank et al., 1980; Vernon and Mior, 1991). NDI also contains 10 different items scored on a 6-point scale, four items from the original ODI (pain intensity, personal care, lifting, sleeping), one new item (recreation) replacing the item “sex-life” in ODI, and five new items adapted to neck patients (reading, headache, concentration, work, driving). Like the ODI, the NDI ranges from 0 to 100 percent with 0 representing no limitation and 100 representing maximal limitation. NDI has been translated into Norwegian by the Mapi Institute and is currently under testing for psychometric properties in Norwegian neck patients (www.mapi-trust.org/). 2.5. Measurement completed by manual therapists The clinicians were asked to rate how they would assess the patient’s prognosis by the following question: Based on your clinical assessment, how likely do you think it is that the patient will have improved significantly in function (daily activities) in eight weeks?

Measurements were reported by patients and by manual therapists. Baseline measures filled in by the patients were completed after inclusion in the study, but before the first consultation by manual therapist. Follow-up was by mail eight weeks after inclusion. The manual therapists filled in assessment of improvement immediately after the first patient consultation, blinded for patient’s answers to the questionnaire.

They responded on a 0e10 numerical rating scale (NRS), ranging from 0 (functional improvement is unlikely) to 10 (functional improvement is certain).

2.4. Measurements completed by patients

3. Statistical analyses

Background descriptive data (age, gender, employment status, pain characteristics (duration of current pain episode (acute ¼ 0e2 weeks, sub-acute ¼ 2e12 weeks, chronic ¼ 3e12 months, chronic > 12 months) and pain at present)) are parts of the ÖMPQ,

Background descriptive data is presented as means with standard deviations (SD) and frequency (%) for counts. The ODI/NDI score eight weeks after the initial consultation with manual therapist was used as a continuous dependent variable in the regression

126

H. Dagfinrud et al. / Manual Therapy 18 (2013) 124e129

analyses. The clinicians’ assessment of the patients’ prognosis for improvement in functional capacity was dichotomised into good prognosis (8e10) and poor prognosis (1e7), based on the distribution of the responses (cut-off according to the median). The continuous ÖMPQ score was used in the regression models, whereas the three categories of the ÖMPQ (low risk for prolonged disability ( 105) were used in the analyses of the instruments’ precision. The ODI/NDI change score was dichotomised into improved (mean change  10 points) and not improved (mean change < 10 points) (Fairbank and Pynsent, 2000; Ostelo et al., 2008). The dichotomised change score was used to estimate the sensitivity, specificity, the Area under Curve (AUC) with 95% and the proportion of false positives and negatives classified by the clinicians’ prognostic assessment and by the ÖMPQ. Further, the likelihood ratios (LRþ and LR) were estimated to describe odds favouring functional improvement given a certain result. LR is estimated by combining sensitivity and specificity (LR: 1-sensitivity/specificity, LRþ: sensitivity/1-specificity), and an LR of one indicates that the test result does nothing to change the odds favouring the condition, whereas an LR greater than one increases the odds of the condition and an LR less than one diminishes the odds favouring the condition. The best test definition for ruling in a condition is the one with the largest positive LR, whereas the best definition for ruling out a condition would have the smallest negative LR (Altman, 1991). Univariate regression analyses were performed to examine the associations between the dependent variable (ODI/NDI end-score) and the explanatory variables (age, gender, duration of pain, ODI/ NDI baseline scores) at eight weeks follow-up. Separate multiple linear regression models for LBP and neck patients were built to explore the explanatory power of the ÖMPQ (model C), the clinicians’ prognostic assessment (model B) and of both tools in the same model (model A), on the ODI/NDI end-score (eight weeks after first consultation). Adjustments were made for age, gender, duration of pain and baseline ODI/NDI scores. 4. Results 157 Patients were included, mean (SD) age 44 (14) years, 81 with neck pain, 76 with LBP (Table 1). The gender ratio was 72% and 46% female in the neck pain and LBP group, respectively (p < 0.001). LBP patients reported significantly higher disability than the neck-pain

group (ODI/NDI: 36 (17)/27 (13), p < 0.001), whereas the ÖMPQ mean score was similar across groups (neck pain vs LBP: 78 (26) vs 84 (30) p ¼ 0.18) (Table 1). 128 Patients completed the ODI/NDI at follow-up 8 weeks after first consultation. The clinicians’ assessment of prognosis of improvement tended to be better for the LBP patients than for neck patients, but the difference was not significant (p ¼ 0.06). In the LBP group, functional improvement (ODI change > 10) could be predicted with 56% specificity and 69% sensitivity with the clinicians’ assessment tool at eight weeks follow-up. By use of the ÖMPQ, the specificity was 78% and the sensitivity 21% (Table 2). The clinicians’ assessment classified 55% false negatives, while the ÖMPQ classified 69% false negatives. In the neck-pain group, the specificity for predicting improvement was 56% for the clinicians’ assessment tool and 86% for the ÖMPQ, while the sensitivity was 55% and 18% respectively. The best odds for predicting the outcome for LBP patients at eight weeks follow-up was found for the clinicians’ assessment of prognosis (LRþ ¼ 2.1 and LR ¼ 0.55), whereas the likelihood ratios were similar for the two methods in the neck group (Table 2). The AUC (95% confidence interval (CI)) for prediction of functional outcome in the LBP patient group was 0.58 (0.42, 0.73) for the ÖMPQ and 0.62 (0.46, 0.78) for the clinicians’ assessment of prognosis. For the neck-pain group, the AUC (95% CI) was 0.60 (0.44, 0.75) for the ÖMPQ and 0.58 (0.42, 0.74) for clinicians’ assessment. The LBP group showed significantly larger functional improvement than the neck patient group (mean (SD) ODI/NDI change scores LBP 20.4 (16.5) vs neck patients 7.4 (11.1), p < 0.001). In the LBP group, the baseline ODI score, the clinicians’ prognostic assessment and the ÖMPQ were significantly associated with the ODI scores 8 weeks after the initial consultation in the univariate analyses (p < 0.01), whereas age, gender and duration of pain were not (p > 0.1). Duration of pain was removed, while age and gender were kept in the final, multivariate model. The R2 of the total model (both instruments in the model) was 0.33 (explaining 33% of the 8-weeks ODI scores). The relative contribution (R2 change) of ODI baseline score, the clinicians’ prognostic assessment and the ÖMPQ was similar, but only the ÖMPQ contributed significantly in this model (p ¼ 0.02), but (Table 3 model A). Further, in the separate models, both the clinicians’ prognostic assessment and the ÖMPQ were significant contributors (R2 9%, p ¼ 0.02 and R2 15%, p ¼ 0.002, respectively) (Table 3, model B and C). Additionally, the ODI baseline score was a significant

Table 1 Baseline characteristics, total cohort and for subgroups of LBP and neck pain. Total group (n ¼ 157)

Neck pain (n ¼ 81)

LBP (n ¼ 76)

Age (mean SD) Gender (% m/f) Payed work (%)

44.3 (14.4) (minemax 18e81) 41/59 65

43.4 (14.4) 28/72 62

45.3 (14.5) 54/46 69

0.43 0.001 0.37

Duration of pain Acute (0e2 weeks) Sub-acute (2e12 w) Chronic (3e12 mon) Chronic > 1 year

23.4% 24.1% 13.9% 38.6%

20.5% 24.1% 16.9% 38.7%

26.7% 24.0% 10.7% 38.7%

0.63

Pain at present (0e10, 10 ¼ worst) ODI/NDI (100 ¼ worst) Clinicians prognostic assessment (% classified as good prognosis)

6.36 (3.54) 31.36 (15.43) 51%

6.53 (3.43) 27.13 (13.1) 46%

6.15 (3.66) 35.9 (16.5) 61%

0.52 10) for LBP and neck-pain patients 8 weeks after the initial consultation with manual therapist. Prediction tool

Sensitivity

Specificity

False negative

False positive

LRc

þLR

AUCd (95% CI)

LBP patients ÖMPQa Clinicians assessmentb

78% 56%

21% 69%

69% 55%

32% 23%

1.01 0.55

0.95 2.1

0.58 (0.42, 0.73) 0.62 (0.46, 0.78)

Neck-pain patients ÖMPQa Clinicians assessmentb

86% 56%

18% 55%

27% 28%

67% 62%

0.95 0.80

1.29 1.25

0.60 (0.44, 0.75) 0.58 (0.42, 0.74)

a b c d

ÖMPQ cut-off 105. Clinicians assessment, cut-off 7. Likelihood ratio. Area under curve.

contributor in the clinicians’ prognostic assessment-model (model B), but not in the ÖMPQ-model (model C) (R2 12%, p ¼ 0.01 and R2 11%, p ¼ 0.15, respectively). In the neck patient group, the baseline NDI score, duration of pain, the clinicians’ prognostic assessment and the ÖMPQ were significantly associated with the NDI scores 8 weeks after the initial consultation in the bivariate analyses (p < 0.01). Age and gender were not significantly associated, but were kept in the final, multivariate model (Table 4). The R2 of the total model (Table 4, model A, both tools included) was 0.64. The NDI baseline scores and duration of pain contributed significantly in the adjusted model (R2 53%, p < 0.001 and R2 8%, p ¼ 0.03 resp), whereas neither the ÖMPQ nor the clinicians’ prognostic assessment was significant explanatory variables in the total model (model A) (p ¼ 0.11 and p ¼ 0.68, resp). The results were similar for the two separate models; the baseline NDI scores and the duration of pain were significant contributors in the models, while the clinicians’ prognostic assessment (p ¼ 0.67, model B) and the ÖMPQ (p ¼ 0.07, model C) were not significantly associated with the dependent variable NDI. Furthermore, age and gender were not significant in any of the models for neck patients (Table 4). The explanatory power was high in all models (R2 0.64, 0.62 and 0.64 resp). 5. Discussion The results of this study indicate that both the ÖMPQ and the clinicians’ prognostic assessment can be used as tools for predicting

functional outcome 8 weeks after the initial assessment of manual therapist for LBP patients, whereas the prediction of outcome for neck patients is unsure. Neither of the two methods showed high precision in their prediction, however. In Norway, LBP patients constitute approximately 27% of the physiotherapists’ treatment, 82% of the chiropractors’ treatments and 5e10% of the general practitioners’ consultations (Werner et al., 2008). A similar picture of health care consumption has been reported for neck-pain patients (Cote et al., 2009; Hogg-Johnson et al., 2009). An early identification of patients at risk of developing longterm problems would allow for allocation of resources to those who are most in need of further attention, and thereby potentially reducing the costs and suffering associated with persistent pain and disability (Linton and Boersma, 2003; Boersma and Linton, 2005). However, predicting outcome in patients with back- and neck pain is challenging due to a number of various and often unknown underlying pathological processes (Hurwitz et al., 2009; Chou and Shekelle, 2010). The ÖMPQ was developed to provide important information about the patient’s experience of his/her condition which might be difficult to catch in a standard medical examination. A potential disadvantage with the ÖMPQ, however, is that it takes some minutes for the clinician to administer and score, probably preventing some clinicians from using it as a screening tool. Therefore, in this study, we aimed at comparing the predictive value of the ÖMPQ with a simple, but explicitly formulated question about the clinicians’ over-all assessment of the patient’s prognosis for functional improvement. The clinicians’ assessment

Table 3 Regression models for the LBP patient group, with ODI score 8 weeks after the initial consultation as dependent variable. Variable

Crude estimates (beta, 95% CI)

R2

p-Value

Age Gender Baseline ODI Duration of pain Clin pred ÖMPQ R2 model A

0.2 0.8 0.3 2.0 2.6 0.2

0.04 0.001 0.13 0.04 0.13 0.21

0.11 0.80 0.01 0.14 0.01